The permeability transition and BCL-2 family proteins in apoptosis: co-conspirators or independent agents?

Programmed cell death (PCD), or apoptosis, is part of the normal biology of all metazoan organism and, when inappropriately activated or inhibited, has been implicated in a variety of human diseases. Cell death pathways critically depend on the participation of mitochondria; indeed, mitochondria play a pivotal role in cell survival and tissue development not only by virtue of their role in apoptosis but also due to their key function in energy production. These roles of mitochondria center, in part, around their participation in the dynamic processes by which cellular levels of Ca²⁺ are modulated. In this context, mitochondria are critical contributors to the regulation of cellular Ca²⁺ levels, accumulating cytoplasmic Ca²⁺ whenever the local cytoplasmic free Ca²⁺ rises above a critical ‘set point’ and slowly releasing Ca²⁺ when cytoplasmic Ca²⁺ is lowered below this set point (for reviews, see Vandecasteele et al¹, Bernardi² and Nicholls and Budd³). In addition, the close coupling of mitochondrial and ER Ca²⁺ levels has defined an interaction that is tightly and subtly organized, and coordinated on a subcellular basis, so that spatial and temporal details of the Ca²⁺ signal may be defined by the spatial organization and possibly specialization of mitochondrial populations within cells (for review, see Rizzuto et al.⁴

In most vertebrates, the pivotal event in the apoptotic process involves the permeabilization of the outer mitochondrial membrane (OMM), resulting in the release of a variety of proteins normally resident in the space between the two mitochondrial membranes (intermitochondrial membrane space; IMS), including proteins that activate caspase-independent apoptotic cascades (e.g., apoptosis-inducing factor) or regulators of caspase-dependent cell death pathways (e.g., cytochrome c (cyt c)), which trigger downstream events that include the formation of the ‘apoptosome’, activation of downstream caspases (the effectors of PCD), and nuclear degradation (for recent review, see Van Loo et al⁵). While the mechanisms leading to the cytosolic release of IMS proteins remains the subject of intense debate, it is clear that many proteins can modulate this process. Most recently, a formal distinction between two classes of mechanisms for OMM permeabilization have been elaborated, reflecting a historical separation of the two processes in question, the permeability transition (PT) and the action of members of the BCL-2 family of proteins.⁶

Inner Membrane Permeability Transition

The first proposed mechanism for OMM permeabilization depends on both the OMM and the inner mitochondrial membrane (IMM) and is based on the fact that the regulation of ion fluxes across the IMM is essential, since energy conservation in the form of a proton electrochemical potential difference (ΔμH) is used to drive each process (for review, see Bernardi et al⁷). The IMM possesses an intrinsically low permeability to ions and solutes, and a set of channels and transporters regulate mitochondrial ion fluxes and volume homeostasis. However, mitochondria in vitro can easily undergo a permeability increase to solutes with molecular masses of about 1500 Da or lower (for reviews, see Crompton et al⁸ and Zoratti and Szabo⁹). This permeability change, called the PT, is widely thought to be driven by the formation of a complex of proteins in the IMM called the PT pore (PTP), and is followed by an osmotically obligatory water flux across the membrane due to the high protein concentration of the mitochondrial matrix. Activation of the PTP, at least in vitro, results in passive swelling of the IMM and rupture of the OMM, with ensuing release of IMS components. Consequently, the PT has dramatic consequences on mitochondrial function, resulting in the collapse of the membrane potential across the inner mitochondrial membrane required to drive the synthesis of ATP by the F0/F1 ATPase, and in the depletion of pyridine nucleotides and respiratory substrates, which eventually causes respiratory inhibition after transient uncoupling.

The molecular composition of the PTP has been, and remains, hotly debated. The inner and outer mitochondrial membranes appear to interact in specialized regions called the ‘contact sites’, which are enriched in adenine nucleotide translocase (ANT) of the IMM and the voltage-dependent anion channel (VDAC) of the OMM bound to cytosolic hexokinase I. Reconstitution of PTP activity after partial purification of these hexokinase-bound complexes led to the idea that the PTP may be formed by ANT and VDAC molecules interacting at these sites,^{10, 11, 12, 13} possibly along with ancillary regulatory proteins such as cyclophilin D (CyP-D), hexokinase and the peripheral benzodiazepine receptor (PBR) (a review of the data supporting this view of the molecular composition of the PTP can be found in Crompton¹⁴). However, conclusive evidence that the ANT is not essential for PTP formation was obtained in a detailed analysis of mitochondria lacking all ANT isoforms, which revealed that a Ca²⁺-dependent PT took place.¹⁵ In addition, since the PT is primarily an IMM event, VDAC, the PBR, and hexokinase must be conservatively viewed as regulators of the formation of the inner membrane PTP. While their involvement in the regulation of the PTP remains reasonable, convincing data directly implicating these protein in pore formation or regulation in an in vivo context are absent. Thus, of the proposed components of the PTP, convincing genetic and biochemical evidence only exists for the role of CyP-D in the regulation of PTP activity. Given these considerations then, we must admit that the proteins actually forming core components of the PTP remain a mystery (see Bernardi et al¹⁶ for a detailed discussion).

Direct Outer Membrane Permeabilization

The second class of mechanism for OMM permeabilization is suggested not to involve a major role for the IMM but rather, is proposed to depend on the action of members of the BCL-2 family of proteins. Genetic analysis in mice has clearly demonstrated that members of this family regulate apoptosis, in part, through their action on mitochondria (for review, see Breckenridge and Xue¹⁷). Indeed, loss of function studies have demonstrated that the absence of key multidomain proapoptotic members of this family, BAX and BAK, creates a profound block in PCD pathways in which mitochondria are preserved after the initiation of apoptotic signals at seemingly unrelated and distinct cellular sites including the plasma membrane, nucleus and ER. While this class of proapoptotic molecules is required for OMM permeabilization, other members of this family (e.g., BCL-2, BCL-X_L) act to prevent OMM permeabilization and hence are antiapoptotic. BH3-only members of this family appear to promote the action of proapoptotic molecules like BAX or BAK, or inhibit the action of antiapoptotic family members like BCL-2 and BCL-X_L. How members of the BCL-2 family accomplish these feats is still not clear. Most theories are based on the observation that multidomain pro- and anti-apoptotic family members can form channels when reconstituted into synthetic lipid bilayers and vesicles. Channels formed by these molecules are generally of indeterminate size that may be large enough to allow the passage of proteins across the OMM. However, the formation of novel channels in the OMM in response to apoptotic signals has yet to be convincingly demonstrated. Thus, the relationship between in vitro channel formation and actual cellular processes occurring during PCD remains unclear.

BCL-2 and the PTP

Historically, the PTP and proteins of the BCL-2 family have been considered to drive mechanistically independent processes. However, early hints of a connection between these events arose from studies indicating that the PTP may be modulated by BCL-2. For example, BCL-2-overexpressing cells were more resistant to PTP induction by uncoupler and tert-butylhydroperoxide but not diamide and Ca²⁺ overload.¹⁸ Biochemical links between components traditionally viewed to form the PTP (e.g., VDAC or the ANT) and BCL-2 family members also strengthened the view that the role of the processes mediated by these two complexes of proteins were in some way coordinated, although many contradictory reports exist in the literature (e.g.^{19, 20, 21, 22, 23}). Later, it was reported that BCL-2 overexpression also inhibited the PTP-dependent release of apoptosis inducing factor from the intermembrane space.²⁴ Intriguingly, mitochondria isolated from rats fed with 2-acetylaminofluorene display a striking upregulation of mitochondrial BCL-2 and desensitization of the PTP, and both events match the increased hepatocyte resistance to cytokine-induced apoptosis in vivo.²⁵ As mentioned above, however, antiapoptotic members of the BCL-2 family apparently exert their stabilizing effects at the OMM by counteracting the channel-forming properties of BAX and BAK. Given that the PT is primarily an IMM event, either BCL-2 family members indirectly affect the PTP through the outer membrane, or they can also act on the inner membrane as suggested by the effects of cyclosporin A (CsA) on BAX-dependent inner membrane remodeling and Ca²⁺-dependent apoptosis (see below).

A further link between these two processes was also uncovered in studies from the Korsmeyer lab demonstrating that the PT may be responsible for the structural reorganization of the mitochondria following activation of pro-apoptotic BCL-2 family members to ensure that cyt c release during PCD is rapid and complete.²⁶ These studies have demonstrated that pro-apoptotic members of the BCL-2 family drive a striking remodelling of mitochondrial structure. During this process, individual cristae become fused and the junctions between cristae and the IMM are opened, resulting in the mobilization of cyt c stores normally restricted to tubular cristae. This process may be independent of the action of proapoptotic BCL-2 family members but is inhibited by CsA, an inhibitor of the PTP, directly implicating the PT in this reorganization.²⁶ Thus, it may be that the apoptotic pathway bifurcates following the activation of proapoptotic BCL-2 family members, resulting in an initial release of 10–15% of the cyt c residing in the IMS²⁷ and consequent caspase activation, while a separate path of mitochondrial remodelling dependent on the PT may eventually lead to matrix swelling with complete release of cyt c and onset of mitochondrial dysfunction.²⁸

Other studies designed to test the relationship of the PT to the action of members of the BCL-2 family of proteins have taken advantage of the fact the PTP can be inhibited by CsA. Mice in which the mitochondrial target of CsA action, CyP-D, has been eliminated by ‘knock out’ strategies have recently been described.^{29, 30, 31, 32} These studies have demonstrated that cells missing CyP-D show enhanced resistance to challenges that may mediate cell death in a PT-dependent manner (e.g., mitochondrial Ca²⁺ overload and responses to reactive oxygen) but not to treatment with agents that act through BCL-2-dependent pathways (e.g., staurosporine and etoposide).

Two studies also assessed the release of cyt c in vitro from mitochondria prepared from the liver of wild-type and CyP-D-null mice following the application of recombinant forms of proapoptotic BCL-2 proteins.^{30, 32} Treatment of CyP-D-null mitochondria with either tBID or BAX demonstrated that CyP-D deficiency had no effect on the ability of these proteins to promote the release of cyt c when compared to wild-type mitochondria. However, the finding that matrix CyP-D does not play a major role in cyt c release by tBID and BAX added to isolated mitochondria in vitro is not surprising, since this protocol results in direct permeabilization of the outer mitochondrial membrane. On the other hand, Baines et al³² showed that CyP-D-deficient cells are not resistant to apoptosis otherwise caused by over expression of BAX. In contrast, and as would be predicted from the cellular studies noted above, Ca²⁺-induced (i.e., PTP-dependent) cyt c release was predictably reduced in CyP-D-deficient mitochondria at Ca²⁺ loads that caused pore opening only in the wild-type mitochondria.

The question of the relationship between the PTP and BCL-2 has more recently been tackled by characterizing the mitochondrial effects of BH3I-2', Chelerythrine, HA14-1 and EM20-25. These are small organic molecules that share the ability to bind the BH3 domain of BCL-2 and thus displace BAX and result in apoptosis.³³ All compounds tested sensitize the PTP to opening in isolated mitochondria and intact cells; EM20-25 caused activation of caspase-9, neutralization of the antiapoptotic activity of overexpressed BCL-2 towards staurosporine, and sensitisation of BCL-2-expressing cells from leukemic patients to the killing effects of staurosporine, chlorambucil and fludarabine.³³

An Integrated Model

Given the comments outlined above, it would seem unnecessary to consider the PTP and the action of BCL-2 proteins as separate and distinct processes. However, if they function coordinately to mediate the permeabilization of the OMM and the release of proapoptotic factors from the IMS, how might their actions be integrated? A critical point would seem to be the involvement of large amplitude swelling of mitochondria during PCD in concert with mitochondrial depolarization, as perhaps mediated by the PTP, a point that is hotly debated. Although common PT assays in vitro are based on swelling, swelling is not a necessary consequence of pore opening in vivo since swelling results from the solute and water flux from the intermembrane/intercristal spaces to the matrix only if there is an osmotic imbalance. Indeed, in vitro, PT-dependent swelling can be completely prevented by tuning the concentration of nondiffusible macromolecules to counterbalance the osmotic pressure developed by the matrix proteins, conditions certainly more closely reflective of those encountered in living cells.³⁴ Yet, PTP-dependent swelling of mitochondria in vivo has been observed in response to insults such as toxic hepatic injury and ischemia damage in the brain and heart, suggesting that at least under these pathological conditions pore opening causes release of apoptogenic proteins through matrix swelling (e.g.^{35, 36}) On the other hand, and predictably for the needs of energy conservation, under normal conditions the inactive state of the PT would be favored, making prolonged opening an uncommon event. Rather, in healthy cells the PT would normally open only transiently (i.e., reversibly) to states that are not associated with mitochondrial swelling or stable changes of the ΔμH. These transient states were originally noted at the single channel level as a flickering of the ‘mitochondrial megachannel’ in the millisecond time scale between lower conductance open and closed states.³⁷ These states may correspond to brief openings of the PTP in situ that could be detected by trapped calcein but not by potentiometric probes.³⁸

It is reasonable then, to suggest that integration of the action of the PT and BCL-2 family members occurs at the level of transient activation of the PTP, since under normal conditions mitochondrial swelling is likely not associated with PTP opening. Indeed, several reports implicate transient activation of the PTP, not associated with mitochondrial swelling, in processes that underlie PCD and action of BCL-2 family members (e.g.³⁹). Furthermore, short PTP opening is likely to mediate fast release of Ca²⁺ from mitochondria.^{28, 38, 40, 41, 42} As Ca²⁺ is the most important cellular permissive factor for PTP opening, the close coupling of mitochondrial and ER Ca²⁺ levels may provide an initial level of coordination between the action of the PTP and BCL-2 family members. A number of studies have demonstrated that members of the BCL-2 family reside not only in the OMM but also in the ER where they have opposing actions in regulating the transfer of ER Ca²⁺ to mitochondria; antiapoptotic members reduce ER Ca²⁺ and pro-apoptotic members promote Ca²⁺ mobilization from the ER to mitochondria during apoptosis, perhaps by regulating of the activity of the ER inositol trisphosphate receptor.^{43, 44, 45} Thus, the action of the PT and BCL-2 family members may be harmonized through the ability of BCL-2 proteins to regulate the levels of ER Ca²⁺, the key trigger of the mitochondrial PT.

Consequently, it is logical to think that these pathways are coordinated by the dynamic processes through which cellular levels of Ca²⁺ are modulated. Indeed, the ability of the PT to regulate, and respond to, mitochondrial Ca²⁺ has already been demonstrated to play a central role in PCD pathways and in the mechanisms by which BCL-2 family members may influence this process.^{46, 47, 48} For example, in cardiac myotubes exposed to apoptotic agents, Ca²⁺ spikes initiate depolarisation of mitochondria in discrete subcellular regions; these mitochondria initiate slow waves of depolarization and Ca²⁺ release propagating through the cell. Travelling mitochondrial waves are prevented by antiapoptotic BCL-2 family members, perhaps by reducing the transfer of ER Ca²⁺ stores to mitochondria, and are likely to involve PTP activation, since they can be blocked by compounds like CsA known to inhibit the PT.⁴⁹ Mitochondrial Ca²⁺ waves result in cyt c release, caspase activation and PCD; mitochondrial Ca²⁺ release through the PT is critical for wave propagation. Remarkably, the process is promoted by the action of proapoptotic members of the BCL-2 family, which transform mitochondria into an excitable state by sensitizing the PT to Ca²⁺.^{49, 50}

The propagation of mitochondrial Ca²⁺ waves through tubular networks of mitochondria may also serve as an additional coordination point for the action of Ca²⁺, the PTP and BCL-2 proteins through events leading to mitochondrial fission (for example, mediated by Drp1), as has been observed in a number of cell types undergoing PCD.^{51, 52, 53} Yet, the role of mitochondrial fragmentation in apoptosis remains to be clearly established since recent work has suggested that mitochondrial fission can occur independently or downstream of OMM permeabilization depending on the specific PCD stimulus (for review, see Perfettini et al⁵⁴). Consequently, fragmentation of mitochondria may inhibit apoptosis in response to a specific set of stimuli, perhaps those dependent on the propagation of mitochondrial Ca²⁺ waves, since disruption of a tubular mitochondrial network would inhibit propagation of these waves.

In summary, while it may be conceptually easier to view the PTP and BCL-2 family members as controlling distinct pathways to the permeabilization of the OMM and ensuing activation of PCD pathways, it is probably more accurate to view these processes as coordinated and integrated, through the participation of mitochondria in the regulation of cellular Ca²⁺ levels. Finally, as should be obvious from the comments presented above, key contributions by Stan Korsmeyer and his colleagues stand out as providing clear answers in an often murky background. This clarity, which characterized Stan's work throughout his impressive career, allows the potential integration of the two processes forming the subject of this narrative and stands in testament to his insight and vision.